4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; if not, write to the
18 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 * Boston, MA 021110-1307, USA
24 * Copyright (c) 2012, 2015, Intel Corporation.
25 * Use is subject to license terms.
27 * Author: Johann Lombardi <johann@whamcloud.com>
30 #include <lustre_quota.h>
32 #include "osd_internal.h"
35 * Helper function to retrieve DMU object id from fid for accounting object
37 inline int osd_quota_fid2dmu(const struct lu_fid *fid, uint64_t *oid)
41 LASSERT(fid_is_acct(fid));
42 switch (fid_oid(fid)) {
44 *oid = DMU_GROUPUSED_OBJECT;
47 *oid = DMU_USERUSED_OBJECT;
57 * Helper function to estimate the number of inodes in use for a give uid/gid
58 * from the block usage
60 static uint64_t osd_objset_user_iused(struct osd_device *osd, uint64_t uidbytes)
62 uint64_t refdbytes, availbytes, usedobjs, availobjs;
63 uint64_t uidobjs, bshift;
65 /* get fresh statfs info */
66 dmu_objset_space(osd->od_os, &refdbytes, &availbytes,
67 &usedobjs, &availobjs);
69 /* estimate the number of objects based on the disk usage */
70 bshift = fls64(osd->od_max_blksz) - 1;
71 uidobjs = osd_objs_count_estimate(refdbytes, usedobjs,
72 uidbytes >> bshift, bshift);
74 /* if we have at least 1 byte, we have at least one dnode ... */
75 uidobjs = max_t(uint64_t, uidobjs, 1);
81 * Space Accounting Management
85 * Return space usage consumed by a given uid or gid.
86 * Block usage is accurrate since it is maintained by DMU itself.
87 * However, DMU does not provide inode accounting, so the #inodes in use
88 * is estimated from the block usage and statfs information.
90 * \param env - is the environment passed by the caller
91 * \param dtobj - is the accounting object
92 * \param dtrec - is the record to fill with space usage information
93 * \param dtkey - is the id the of the user or group for which we would
94 * like to access disk usage.
96 * \retval +ve - success : exact match
97 * \retval -ve - failure
99 static int osd_acct_index_lookup(const struct lu_env *env,
100 struct dt_object *dtobj,
101 struct dt_rec *dtrec,
102 const struct dt_key *dtkey)
104 struct osd_thread_info *info = osd_oti_get(env);
105 char *buf = info->oti_buf;
106 size_t buflen = sizeof(info->oti_buf);
107 struct lquota_acct_rec *rec = (struct lquota_acct_rec *)dtrec;
108 struct osd_object *obj = osd_dt_obj(dtobj);
109 struct osd_device *osd = osd_obj2dev(obj);
114 rec->bspace = rec->ispace = 0;
116 /* convert the 64-bit uid/gid into a string */
117 snprintf(buf, buflen, "%llx", *((__u64 *)dtkey));
118 /* fetch DMU object ID (DMU_USERUSED_OBJECT/DMU_GROUPUSED_OBJECT) to be
120 rc = osd_quota_fid2dmu(lu_object_fid(&dtobj->do_lu), &oid);
124 /* disk usage (in bytes) is maintained by DMU.
125 * DMU_USERUSED_OBJECT/DMU_GROUPUSED_OBJECT are special objects which
126 * not associated with any dmu_but_t (see dnode_special_open()). */
127 rc = zap_lookup(osd->od_os, oid, buf, sizeof(uint64_t), 1,
130 /* user/group has not created anything yet */
131 CDEBUG(D_QUOTA, "%s: id %s not found in DMU accounting ZAP\n",
132 osd->od_svname, buf);
137 if (!osd_dmu_userobj_accounting_available(osd)) {
138 if (rec->bspace != 0)
139 /* estimate #inodes in use */
140 rec->ispace = osd_objset_user_iused(osd, rec->bspace);
143 snprintf(buf, buflen, OSD_DMU_USEROBJ_PREFIX "%llx",
145 rc = zap_lookup(osd->od_os, oid, buf, sizeof(uint64_t), 1,
149 "%s: id %s not found dnode accounting\n",
150 osd->od_svname, buf);
151 } else if (rc == 0) {
160 * Initialize osd Iterator for given osd index object.
162 * \param dt - osd index object
163 * \param attr - not used
165 static struct dt_it *osd_it_acct_init(const struct lu_env *env,
166 struct dt_object *dt,
169 struct osd_thread_info *info = osd_oti_get(env);
170 struct osd_it_quota *it;
171 struct lu_object *lo = &dt->do_lu;
172 struct osd_device *osd = osd_dev(lo->lo_dev);
176 LASSERT(lu_object_exists(lo));
179 RETURN(ERR_PTR(-ENOMEM));
183 RETURN(ERR_PTR(-ENOMEM));
185 memset(it, 0, sizeof(*it));
186 rc = osd_quota_fid2dmu(lu_object_fid(lo), &it->oiq_oid);
190 /* initialize zap cursor */
191 rc = osd_zap_cursor_init(&it->oiq_zc, osd->od_os, it->oiq_oid, 0);
197 /* take object reference */
199 it->oiq_obj = osd_dt_obj(dt);
202 RETURN((struct dt_it *)it);
206 * Free given iterator.
208 * \param di - osd iterator
210 static void osd_it_acct_fini(const struct lu_env *env, struct dt_it *di)
212 struct osd_it_quota *it = (struct osd_it_quota *)di;
215 osd_zap_cursor_fini(it->oiq_zc);
216 osd_object_put(env, it->oiq_obj);
223 * Move on to the next valid entry.
225 * \param di - osd iterator
227 * \retval +ve - iterator reached the end
228 * \retval 0 - iterator has not reached the end yet
229 * \retval -ve - unexpected failure
231 static int osd_it_acct_next(const struct lu_env *env, struct dt_it *di)
233 struct osd_it_quota *it = (struct osd_it_quota *)di;
234 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
238 if (it->oiq_reset == 0)
239 zap_cursor_advance(it->oiq_zc);
241 rc = -zap_cursor_retrieve(it->oiq_zc, za);
242 if (rc == -ENOENT) /* reached the end */
248 * Return pointer to the key under iterator.
250 * \param di - osd iterator
252 static struct dt_key *osd_it_acct_key(const struct lu_env *env,
253 const struct dt_it *di)
255 struct osd_it_quota *it = (struct osd_it_quota *)di;
256 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
261 rc = -zap_cursor_retrieve(it->oiq_zc, za);
264 rc = kstrtoull(za->za_name, 16, &it->oiq_id);
266 RETURN((struct dt_key *) &it->oiq_id);
270 * Return size of key under iterator (in bytes)
272 * \param di - osd iterator
274 static int osd_it_acct_key_size(const struct lu_env *env,
275 const struct dt_it *di)
278 RETURN((int)sizeof(uint64_t));
282 * zap_cursor_retrieve read from current record.
283 * to read bytes we need to call zap_lookup explicitly.
285 static int osd_zap_cursor_retrieve_value(const struct lu_env *env,
286 zap_cursor_t *zc, char *buf,
287 int buf_size, int *bytes_read)
289 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
292 rc = -zap_cursor_retrieve(zc, za);
293 if (unlikely(rc != 0))
296 if (unlikely(za->za_integer_length <= 0))
299 actual_size = za->za_integer_length * za->za_num_integers;
301 if (actual_size > buf_size) {
302 actual_size = buf_size;
303 buf_size = actual_size / za->za_integer_length;
305 buf_size = za->za_num_integers;
308 rc = -zap_lookup(zc->zc_objset, zc->zc_zapobj,
309 za->za_name, za->za_integer_length,
313 *bytes_read = actual_size;
319 * Return pointer to the record under iterator.
321 * \param di - osd iterator
322 * \param attr - not used
324 static int osd_it_acct_rec(const struct lu_env *env,
325 const struct dt_it *di,
326 struct dt_rec *dtrec, __u32 attr)
328 struct osd_thread_info *info = osd_oti_get(env);
329 zap_attribute_t *za = &info->oti_za;
330 struct osd_it_quota *it = (struct osd_it_quota *)di;
331 struct lquota_acct_rec *rec = (struct lquota_acct_rec *)dtrec;
332 struct osd_object *obj = it->oiq_obj;
333 struct osd_device *osd = osd_obj2dev(obj);
339 rec->ispace = rec->bspace = 0;
341 /* retrieve block usage from the DMU accounting object */
342 rc = osd_zap_cursor_retrieve_value(env, it->oiq_zc,
343 (char *)&rec->bspace,
344 sizeof(uint64_t), &bytes_read);
348 if (!osd_dmu_userobj_accounting_available(osd)) {
349 if (rec->bspace != 0)
350 /* estimate #inodes in use */
351 rec->ispace = osd_objset_user_iused(osd, rec->bspace);
355 /* retrieve key associated with the current cursor */
356 rc = -zap_cursor_retrieve(it->oiq_zc, za);
357 if (unlikely(rc != 0))
360 /* inode accounting is not maintained by DMU, so we use our own ZAP to
361 * track inode usage */
362 rc = -zap_lookup(osd->od_os, it->oiq_obj->oo_dn->dn_object,
363 za->za_name, sizeof(uint64_t), 1, &rec->ispace);
365 /* user/group has not created any file yet */
366 CDEBUG(D_QUOTA, "%s: id %s not found in accounting ZAP\n",
367 osd->od_svname, za->za_name);
375 * Returns cookie for current Iterator position.
377 * \param di - osd iterator
379 static __u64 osd_it_acct_store(const struct lu_env *env,
380 const struct dt_it *di)
382 struct osd_it_quota *it = (struct osd_it_quota *)di;
385 RETURN(osd_zap_cursor_serialize(it->oiq_zc));
389 * Restore iterator from cookie. if the \a hash isn't found,
390 * restore the first valid record.
392 * \param di - osd iterator
393 * \param hash - iterator location cookie
395 * \retval +ve - di points to exact matched key
396 * \retval 0 - di points to the first valid record
397 * \retval -ve - failure
399 static int osd_it_acct_load(const struct lu_env *env,
400 const struct dt_it *di, __u64 hash)
402 struct osd_it_quota *it = (struct osd_it_quota *)di;
403 struct osd_device *osd = osd_obj2dev(it->oiq_obj);
404 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
409 /* create new cursor pointing to the new hash */
410 rc = osd_zap_cursor_init(&zc, osd->od_os, it->oiq_oid, hash);
413 osd_zap_cursor_fini(it->oiq_zc);
417 rc = -zap_cursor_retrieve(it->oiq_zc, za);
420 else if (rc == -ENOENT)
427 * Move Iterator to record specified by \a key, if the \a key isn't found,
428 * move to the first valid record.
430 * \param di - osd iterator
431 * \param key - uid or gid
433 * \retval +ve - di points to exact matched key
434 * \retval 0 - di points to the first valid record
435 * \retval -ve - failure
437 static int osd_it_acct_get(const struct lu_env *env, struct dt_it *di,
438 const struct dt_key *key)
442 /* XXX: like osd_zap_it_get(), API is currently broken */
443 LASSERT(*((__u64 *)key) == 0);
445 RETURN(osd_it_acct_load(env, di, 0));
451 * \param di - osd iterator
453 static void osd_it_acct_put(const struct lu_env *env, struct dt_it *di)
458 * Index and Iterator operations for accounting objects
460 const struct dt_index_operations osd_acct_index_ops = {
461 .dio_lookup = osd_acct_index_lookup,
463 .init = osd_it_acct_init,
464 .fini = osd_it_acct_fini,
465 .get = osd_it_acct_get,
466 .put = osd_it_acct_put,
467 .next = osd_it_acct_next,
468 .key = osd_it_acct_key,
469 .key_size = osd_it_acct_key_size,
470 .rec = osd_it_acct_rec,
471 .store = osd_it_acct_store,
472 .load = osd_it_acct_load
477 * Quota Enforcement Management
481 * Wrapper for qsd_op_begin().
483 * \param env - the environment passed by the caller
484 * \param osd - is the osd_device
485 * \param uid - user id of the inode
486 * \param gid - group id of the inode
487 * \param space - how many blocks/inodes will be consumed/released
488 * \param oh - osd transaction handle
489 * \param is_blk - block quota or inode quota?
490 * \param flags - if the operation is write, return no user quota, no
491 * group quota, or sync commit flags to the caller
492 * \param force - set to 1 when changes are performed by root user and thus
493 * can't failed with EDQUOT
495 * \retval 0 - success
496 * \retval -ve - failure
498 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
499 qid_t uid, qid_t gid, long long space,
500 struct osd_thandle *oh, bool is_blk, int *flags,
503 struct osd_thread_info *info = osd_oti_get(env);
504 struct lquota_id_info *qi = &info->oti_qi;
505 struct qsd_instance *qsd = osd->od_quota_slave;
506 int rcu, rcg; /* user & group rc */
509 if (unlikely(qsd == NULL))
510 /* quota slave instance hasn't been allocated yet */
513 /* let's start with user quota */
514 qi->lqi_id.qid_uid = uid;
515 qi->lqi_type = USRQUOTA;
516 qi->lqi_space = space;
517 qi->lqi_is_blk = is_blk;
518 rcu = qsd_op_begin(env, qsd, &oh->ot_quota_trans, qi, flags);
520 if (force && (rcu == -EDQUOT || rcu == -EINPROGRESS))
521 /* ignore EDQUOT & EINPROGRESS when changes are done by root */
524 /* For non-fatal error, we want to continue to get the noquota flags
525 * for group id. This is only for commit write, which has @flags passed
526 * in. See osd_declare_write_commit().
527 * When force is set to true, we also want to proceed with the gid */
528 if (rcu && (rcu != -EDQUOT || flags == NULL))
531 /* and now group quota */
532 qi->lqi_id.qid_gid = gid;
533 qi->lqi_type = GRPQUOTA;
534 rcg = qsd_op_begin(env, qsd, &oh->ot_quota_trans, qi, flags);
536 if (force && (rcg == -EDQUOT || rcg == -EINPROGRESS))
537 /* as before, ignore EDQUOT & EINPROGRESS for root */
540 RETURN(rcu ? rcu : rcg);